Open jrieken opened 3 months ago
Hi there 👋 That is a WASM out-of-memory error, which makes sense since you only get the error when using the unquantized (fp32) version of the model that is ~1.34GB in size. If you would like to use the unquantized version, I would recommend you use it with the WebGPU execution provider, which you can try with the (experimental) v3 branch: https://github.com/xenova/transformers.js/pull/545. See the first comment for installation and usage instructions.
System Info
"@xenova/transformers": "^2.17.2"
"onnxruntime-web": "1.18.0",
Environment/Platform
Description
Receiving the
RangeError
when disabling quantizationReproduction
quantized
totrue